Centromere detector and method for determining radiation exposure from chromosome abnormalities

1. A method for determining radiation exposure from chromosome abnormalities present in a specimen comprising the steps of: (a) determining a location of at least one centromere of each chromosome in a cell in an image of a metaphase nucleus by: (i) segmenting an accurately drawn chromosome centerline of each of the chromosomes in the cell, (ii) selecting a longitudinal cross-section of a chromosome image with a minimum width, intensity or a combination of features derived from width and intensity, wherein in a dicentric chromosome, a location of a first centromere is masked to identify a subsequent longitudinal cross-section with a second lowest minimum of combined width; (b) counting the number of centromeres in each chromosome in said cell; (c) computing a computed frequency of dicentric chromosomes in a plurality of said cells; and (d) determining a radiation dose by comparing the computed frequency of dicentric chromosomes with a previously determined dose-response curve from a calibrated source of ionizing radiation.

2. The method of claim 1 , where the location of a first centromere is masked to identify a subsequent longitudinal cross-section of a second centromere with a second lowest minimum of combined width and intensity.

3. The method of claim 1 or 2 , where the image of the metaphase nucleus is selected which has at least one of the following properties: (i) complete chromosome number, (ii) minimize a degree of overlap and intersection of chromosomes, and (iii) the chromosomes exhibit defined boundaries.

4. The method of claim 1 or 2 , where the centromere location is determined from the cross-sectional intensity and/or width profiles of the chromosome perpendicular to the chromosome centerline.

5. The method of claim 1 or 2 , where the chromosome centerline is derived by discrete curve evolution based skeleton pruning.

6. A method of determining a radiation dose comprising the steps of: scanning a plurality of sectors in a predetermined area of a microscope slide for at least one sector containing image content similar in morphology to a metaphase cell nucleus containing discrete, adjacent chromosomes thereby creating an image sector; identifying metaphase chromosomes and portions thereof in the image content of the image sector; determining local classification parameters of metaphase chromosomes and chromosomes connected to neighboring chromosomes; extracting features from the image sector of metaphase chromosomes and chromosomes connected to neighboring chromosomes; applying a classification system based on the extracted chromosome features to separate a plurality of different categories of image sectors, the categories consisting of a nice classification, an overlapped classification, and an overspread classification, wherein a nice classification is an image sector having at least one chromosome that is fully contained in the image sector and having substantially no overlap with another chromosome, an overlapped classification is an image sector having at least two chromosomes that are superimposed on one another, and an overspread classification is an image sector having at least one chromosome that is only partially contained in the image sector; determining at least once local parameter for each image sector; applying a ranking system based on a plurality of global parameters; selecting the image sector having at least one chromosome based on the ranking system wherein the selected image sector has a high rank; determining at least one contour of the at least one chromosome; determining a centerline of the at least one chromosome; locating a most likely position of at least one centromere in the at least one chromosome, whereby a plurality of chromosomes in a cell are classified based on position; measuring a centromere confidence value to measure a probable location of the at least one centromere; masking the most likely location for a first centromere and recomputing a centromere confidence value to determine if subsequent centromeres are present on the chromosome; counting the number of centromeres in the chromosome; computing a computed frequency of dicentric chromosomes in at least one cell; and determining a radiation dose by comparing the computed frequency of each dicentric chromosome with a previously determined dose-response curve from a calibrated source.

7. The method of claim 6 , wherein the global classification parameters comprise at least one parameter selected from a group consisting of an average area, a standard deviation of the area, length, perimeter, aspect ratio, pairwise distance between each chromosome, and combinations thereof.

8. The method of claim 6 , wherein the local ranking parameters comprise at least one parameter selected from a group consisting of total area, ratio of severe overlaps, mean pair-wise distance, brightness, contrast, and combinations thereof.

9. The method of claim 6 , wherein the step of selecting an image sector comprises providing an algorithm using a fuzzy logic system to detect an abnormal dicentric chromosome.

10. The method of claim 6 , wherein determining the contour comprises the steps of: a) pre-processing a fluorescent metaphase image of a 4′,6-diamidino-2-phenylindole (DAPI)-stained chromosome for chromosome segmentation; b) observing an intensity variation of the chromosome; c) performing gray scale conversion; d) extracting a rectangular shaped window that includes the chromosome and an adjacent background, whereby variations in lighting across the image were alleviated; e) normalizing intensities by window center adjustment and mapping across a complete spectrum of possible pixel values; f) applying a threshold based on Otsu's method, whereby a binary image is generated; g) subjecting the binary image to the content and classification-based ranking algorithm; h) applying a scaling factor to the binary image; i) extracting the chromosome from the image by connected component labeling of a 4-connected graph, whereby discontinuities were removed within and at the boundary of the chromosome; j) tracing the initial chromosome contour using a 3×3 neighborhood; and k) applying an active contour to determine the chromosome outline, whereby the active contour iteratively converges towards a closest local minima of image gradients from an initial set of control points.

11. The method of claim 10 , further including an iterated Gradient Vector Flow (GVF) active contour to smooth a boundary map of the binary image.

12. The method of claim 6 , wherein defining the contour comprises obtaining a thickness profile and a width profile of the chromosome using an intensity integrated Laplacian thickness measurement, whereby a Laplacian operator (Δ) yields the second order derivative of the image thereby emphasizing the chromosome contour information.

13. The method of claim 6 , wherein defining the contour comprises using a method of image intensity integration, whereby the image is segmented into a plurality of regions or at least one edge fragment is detected by using a concept of light flux per unit area.

14. The method of claim 6 , wherein determining the centerline of the chromosome comprises a skeleton pruning method based on a discrete curve evolution calculation to produce a pruned skeleton, comprising the steps of: a) modeling the skeleton as a polygon with a predetermined number of vertices, b) partitioning at least one discrete curve evolution contour into polygonal sections, c) removing a plurality of skeletal points that generate at least one point on the polygonal section, whereby the resulting skeleton will have a single spurious branch, d) providing a curve fitting method, and e) sampling control points from the pruned skeleton.

15. The method of claim 6 , wherein the step of locating a most likely position of each centromere comprises the steps of: a) using an end pruned centerline as a reference, whereby the end pruned centerline excludes at least one maximum end of the chromosome, b) drawing at least one trellis line segment perpendicular to the pruned centerline segment at unit length intervals, whereby a predetermined cross sectional is a length of the trellis line segment, c) weighting at least one sample intensity of the trellis line segment with a Gaussian function, whereby cancelling a plurality of image noise and a plurality of undesirable effects introduced by chromosome bending, d) determining the lengths of the trellis line segments from at least one binary result obtained through a Gradient Vector Flow (GVF) binary image, wherein the GVF binary image is a Gaussian kernel active contour, whereby use of the GVF binary image provides enhanced edge characteristics of the centromere, and e) locating the centromere by using Wp, wherein Wp is a width profile of the chromosome, along the trellis line segment on a medial axis of the GVF binary image and Ip, wherein Ip is an intensity profile obtained by getting the weighted average of intensity values of a 4′,6-diamidino-2-phenylindole (DAPI)-stained chromosome image based on a Gaussian function along the GVF binary image limited trellis line segment.

16. The method of claim 6 , wherein the step of locating a most likely position of each centromere comprises the steps of: a) combining a chromosome width with a centerline pixel intensity, and b) using a thickness-based measurement alone, whereby a centromere is accurately detected.

17. The method of claim 6 , wherein an algorithm using a fuzzy logic system calculates an optimal position for each centromere, the algorithm comprising the steps of: a) obtaining a set of line segments covering multiple angles with an angle offset of 2.5° for each centerline pixel in the vicinity of the detected centromere location; b) selecting the two shortest line segments from the set of line segments, thereby creating two candidate sets, wherein a first candidate set of the two candidate sets comprises line segments with the shortest value and a second candidate set of the two candidate sets comprises the second shortest line segments; c) determining ResDA wherein ResDA is an angular difference between a direction of a line segment and an expected direction from the centromere for the two candidate sets; d) calculating an offset value for a plurality of candidate points, wherein the offset value is a pixel interval between the detected centromere location and the centerline pixel location; e) obtaining a fuzzy output result for the two candidate sets; f) selecting a maximum result from the two candidate offset sets; g) selecting the candidate location with the highest fuzzy logic output for each candidate data set; h) calculating a difference between the 2 highest fuzzy logic output values; and i) selecting a new centromere location wherein if the difference is greater than 0.1, the centerline pixel with the highest fuzzy logic output is selected as the new centromere location and if the difference is lower than 0.1, the centerline pixel with the lowest offset from the original location is selected as the new centromere location.

18. The method of claim 9 wherein the algorithm using a fuzzy logic system calculates an optimal position for the centromere comprising the steps of: (a) defining a search space of the image that includes only a desired centromere location, (b) selecting a set of centerline points for the search space based on the desired centromere location, (c) drawing a line segment about every 2.5 degrees through each of the selected centerline points, (d) determining the endpoints of the line segments using binary object contour, (e) determining ResW for each line segment wherein ResW is a ratio of the width of a line segment to an average width of a width profile, (f) identifying two line segments with the lowest ResW value, and (g) applying an algorithm based on a fuzzy logic system to calculate an optimal position for the centromere from the two line segments with the lowest ResW value.

19. A method of detecting at least one abnormal chromosome of claim 6 further comprising the steps of: a) obtaining a thickness profile of each chromosome, b) obtaining a width profile of each chromosome, c) providing an intensity integrated Laplacian thickness measurement, and d) providing a Laplacian operator (Δ) to yield a second order derivative of a chromosome metaphase image to emphasize a chromosome contour, and e) detecting a sister chromatid separation using Orthogonal function representation which analyses partitioned contour shapes at a telomere.

20. A computer-implemented method of detecting at least one dicentric chromosome comprising the steps of: a) detecting a chromosome that does not overlap another chromosome to obtain a non-overlapping chromosome, b) segmenting the non-overlapping chromosome, c) determining a centerline for the chromosome, d) minimizing a width and an intensity of the chromosome, e) detecting a centromere, f) detecting a second centromere location by masking the neighborhood of the first centromere location and then finding the global minima of width and intensity, g) detecting sister chromatid separation, whereby a centromere refinement method is used if the sister chromatid separation is detected, and h) detecting a second centromere.

21. The computer-implemented method of claim 20 , wherein the centromere refinement method is used to process images of metaphase chromosomes prepared by culturing cells with either high concentrations or extended duration exposure to a microtubule inhibitory compound.

22. The computer-implemented method of claim 20 , wherein the centromere refinement method is used to process images of metaphase chromosomes prepared by culturing cells with either low concentrations or shorter duration exposure to a microtubule inhibitory compound.

23. A method of detecting at least one abnormal chromosome of claim 6 or 19 further comprising the steps of: producing image data of centromeres to be recognized by fuzzy logic system application; inputting said image data into a learning vector quantization network to produce optimized moment invariant vectors associated with each of said fuzzy logic system; accessing an algorithm to produce a plurality of candidate chromosome centerlines representing said parameters for said fuzzy logic system; transforming each of said plurality of candidate chromosome centerlines into said parameters for said fuzzy logic system; importing said parameters into said fuzzy logic system for said each of said plurality of candidate chromosome centerlines; simulating said fuzzy logic system with respect to said each of said plurality of candidate chromosome centerlines; inputting said optimized moment invariant vectors into said fuzzy logic system; with respect to said each of said plurality of candidate chromosome centerlines, determining how many of said centromeres are correctly recognized by said fuzzy logic system; selecting one of said plurality of candidate chromosome centerlines if said one of said plurality of candidate chromosome centerlines correctly recognizes all of said centromeres; associating a score with said each of said plurality of candidate chromosome centerlines, said score indicating how many of said centromeres were recognized by said each of said plurality of candidate chromosome centerlines; selecting a percentage of said plurality of candidate chromosome centerlines having better scores; applying a crossover process between said selected percentage of said plurality of candidate chromosome centerlines to produce one or more children; replacing one or more of said plurality of said candidate chromosome centerlines having worst scores with said one or more children, resulting in a new population of candidate chromosome centerline; transforming said new population into fuzzy logic parameters; importing said fuzzy logic parameters of said new population into said fuzzy logic system; simulating said fuzzy logic system with respect to said new population; inputting said optimized moment invariant vectors into said fuzzy logic system; with respect to each of said plurality of candidate chromosome centerlines in said new population, determining how many of said centromeres are correctly recognized by said fuzzy logic system; and selecting one of said plurality of candidate chromosome centerlines in said new population if said one of said plurality of candidate chromosome centerlines in said new population correctly recognizes all of said centromeres.

24. The method as recited in claim 23 , wherein said rules comprise: If ResW is very low and ResDA is low, then output level is ultra high, If ResW is very low and ResDA is medium, then output level is ultra high, If ResW is low and ResDA is low, then output level is very high, If ResW is low and ResDA is medium, then output level is medium high, If ResW is low and ResDA is high, then output level is medium low, If ResW is medium and ResDA is low and ResInt is high then output level is medium high, If ResW is medium and ResDA is medium and ResInt is high then output level is medium low, If ResW is medium and ResDA is high, then output level is very low, If ResW is medium and ResDA is low and ResInt is low then output level is medium low, If ResW is medium and ResDA is medium and ResInt is low then output level is very low, and If ResW is high then output level is ultra low; wherein ResW is a ratio of the width of a line segment to an average width of a width profile, ResDA is an angular difference between a direction of a line segment and an expected direction given by a line perpendicular to P 1 -P 3 through P 2 , wherein P 1 , P 2 and P 3 are any 3 consecutive points on a pruned chromosome centerline and ResInt is an average pixel intensity in a 5 pixel region surrounding a candidate centerline pixel wherein said average pixel intensity value is normalized by a ratio of average pixel intensity local neighborhood to maximum average intensity in a search space.

25. At least one non-transitory computer readable medium that stores a set of instructions for running on a computer system, comprising: a. instructions for receiving one or more electronic files of images into one or more memory; b. instructions for identifying image content resembling a metaphase nucleus containing discrete, adjacent chromosomes; c. instructions for identifying metaphase chromosomes and portions thereof in the image content; d. instructions for determining local classification parameters of metaphase chromosomes and chromosomes connected to neighboring chromosomes; e. instructions for extracting chromosome features from the image content; f. instructions for applying a classification system based on extracted chromosome features to the image content to separate a plurality of different categories of images, the categories consisting of a nice classification, an overlapped classification, and an overspread classification, wherein a nice classification is an image having at least one chromosome that is fully contained in the image and having substantially no overlap with another chromosome, an overlapped classification is an image having at least two chromosomes that are superimposed on one another, and an overspread classification is an image having at least one chromosome that is only partially contained in the image sector; g. instructions for applying a ranking system to the images based upon a plurality of global parameters; h. instructions for selecting a subset of images based upon rank; i. instructions for determining contours of the chromosome; j. instructions for determining the location of the centerline of the chromosome; k. instructions for locating the most likely position of at least one centromere in the chromosome, whereby a plurality of chromosomes in the same cell are classified based on position; l. instructions for measuring a centromere confidence value to measure the probable location of the automated centromere detection process; m. instructions for masking the most likely location for the first centromere and recomputing the centromere confidence value to determine if subsequent centromeres are present on the same chromosome n. instructions for counting the number of centromeres in the chromosome; o. instructions for computing a frequency of dicentric chromosomes in at least one cell; and p. instructions for determining a radiation dose by comparing the computed frequency of each dicentric chromosome with a previously determined dose-response curve from a calibrated source.

26. The set of instructions stored on the at least one non-transitory computer readable medium of claim 25 , wherein the instructions a.-p. are adapted to be executed without manual intervention.

27. The set of instructions stored on the at least one non-transitory computer readable medium of claim 25 , further comprising at least one instruction for storing information indicative of the content resembling a metaphase chromosome.

28. The set of instructions stored on the at least one non-transitory computer readable medium of claim 25 , further comprising at least one instruction for storing information for determining local classification parameters of metaphase chromosomes and chromosomes connected to neighboring chromosomes.

29. The set of instructions stored on the at least one non-transitory computer readable medium of claim 25 , further comprising at least one instruction for storing global ranking parameters of metaphase chromosomes and chromosomes connected to neighboring chromosomes.

30. The set of instructions stored on the at least one non-transitory computer readable medium of claim 25 , further comprising at least one instruction for storing the ranked images.

31. The set of instructions stored on the at least one non-transitory computer readable medium of claim 25 , further comprising at least one instruction for storing contours of chromosomes.

32. The set of instructions stored on the at least one non-transitory computer readable medium of claim 25 , further comprising at least one instruction for storing the location of the centerline of metaphase chromosomes.

33. The set of instructions stored on the at least one non-transitory computer readable medium of claim 25 , further comprising at least one instruction for storing the number of centromeres in a metaphase chromosome.

34. The set of instructions stored on the at least one non-transitory computer readable medium of claim 25 , further comprising at least one instruction for storing the location of centromeres in a metaphase chromosome.

35. The set of instructions stored on the at least one non-transitory computer readable medium of claim 25 , further comprising at least one instruction for storing a determination of an abnormal chromosome.

36. The set of instructions stored on the at least one non-transitory computer readable medium of claim 25 , where the instructions are adapted to determining the frequency of dicentric chromosomes in a set of images of metaphase nuclei.

37. The set of instructions stored on the at least one non-transitory computer readable medium of claim 25 , where the instructions are adapted to determining the centromere confidence value to determine the most probable locations of one or more centromeres in a chromosome.

38. The set of instructions stored on the at least one non-transitory computer readable medium of claim 25 , where the instructions are adapted to determining a radiation dose by comparing the computed frequency of each dicentric chromosome with a previously determined dose-response curve from a calibrated source.